Data for a UserR! 2016 talk
This repository contains a UseR! talk that tries to show how Julia achieves performance. This is done by comparing an implementation of the colMeans function in Julia.
colMeans(), in R, calls the do_colsums() function written in C. To compute 1000 invocations of colMeans on a 100,000 element complex Vector takes 2 secs on my underpowered MacBook.
Implemented in pure R, and using the optimization level 3 of the GNU R bytecode compiler, the comptuation takes about 30 secs.
A straightforward Julia implementation takes .2 seconds.
TODO:
-
fix bug in the colMeans code, for complex, res is dropping values
-
fix bug in the typmin for Complex, NaN can't be compared